Confounder Variable Flagging System — Free Checklist Flagger

This tool screens candidate variables for confounding using the classic three-question checklist. A confounder is a variable associated with the exposure, associated with the outcome, and not itself caused by the exposure (not on the causal pathway). You answer those three questions per variable, and the tool flags each one: likely confounder (adjust for it), mediator (on the pathway — adjusting naively can erase the very effect you study), partially associated (review), or unlikely. Confounding is the main threat to observational studies: without it handled, a treatment can look effective merely because healthier people received it.

How to use it

List each candidate variable with three yes/no answers.
Answer honestly: associated with the exposure? With the outcome? On the causal pathway?
Click Flag variables.
Treat likely confounders as adjustment candidates.
Draw a causal diagram to confirm the flags before analysis.

Worked example

A study asks whether coffee drinking relates to heart disease. Age is associated with both coffee habits and heart risk, and age is not caused by coffee, so age flags as a likely confounder — adjust for it. Blood pressure is associated with both but sits on the pathway (coffee may raise blood pressure, which raises risk), so it flags as a mediator: adjusting for it would hide part of coffee’s effect. Favourite colour is associated with neither and flags as unlikely. Smoking, associated with both and not caused by coffee, flags as a confounder. Four variables, four different treatments — and the checklist took one minute.

Honest limits

A checklist is a screen, not a causal proof. “Associated with” is a judgment call that needs subject knowledge, and the same variable can be a confounder in one study and a mediator in another depending on timing and direction. The tool cannot detect unmeasured confounding at all: if the true confounder was never collected, no checklist finds it, which is why sensitivity analyses matter. It also cannot handle colliders — variables caused by both exposure and outcome — where adjusting creates bias instead of removing it. For a publishable analysis, draw a directed acyclic graph (DAG) with a subject expert and let the graph, not the checklist alone, decide the adjustment set.

Who it is for

This tool is for students learning confounding, researchers planning an observational analysis, and reviewers checking whether a paper adjusted for the right variables. Use it at the design stage, when the adjustment set can still be chosen freely, and revisit it after data collection when associations can be checked empirically.

Going further

The most dangerous adjustment error is over-adjustment: controlling for a mediator, a collider, or a variable measured after the exposure can bias results more than leaving a real confounder uncontrolled. A good habit is to write one sentence per adjustment variable justifying it — “age precedes both coffee habits and heart risk and relates to both” — and one sentence per excluded variable explaining why it stays out. If you cannot write the justification, the variable does not belong in the model yet. Pre-specify the adjustment set in the protocol; adding variables after seeing which ones change the result is a quiet form of result-shopping that reviewers increasingly spot. The checklist starts the conversation, but the written justification is what makes the analysis defensible. A related trap is adjusting for variables measured after the exposure, such as a post-treatment symptom score. Such variables may be mediators or colliders, and controlling for them can manufacture bias where none existed. The timing rule is simple and powerful: an adjustment variable should precede the exposure in time, or at least be clearly unaffected by it. When in doubt about timing, leave the variable out of the primary model and show it in a sensitivity analysis instead — readers can then see both answers and judge which timing story they believe.

Embed this tool on your website — free

Copy the code below to add this working tool to your own page. A small credit link stays with it — that’s all we ask.